FHIN®

[Federated Health Innovation Network]

FHIN® is an open-source/open-science network of hospitals, working on a fully automated, federated, drift-free and FAIR data platform with a guarantee of context and quality, ready to solve concrete questions on a project basis and to operationalise these solutions in practice.

Objective

FHIN® vzw (Federated Health Information Network) enables data-driven medicine and care through a high-performance multicentric data platform, targeting the following goals;

Standardization on multiple levels with unified tooling and standardize the structure and semantics of clinical data & bias reduction through collaboration
Drift-free and FAIR data exchange for enhanced patient outcomes
Quality assurance
Efficient multicentric data platform
Equal partnership to foster trust among 9 partners
Alignment with certification and MDR compliance
Internationally scalable network

For more information on FHIN vzw’s mission, vision, and values, please visit the FHIN website.

Methodology

The FHIN® project will complete the following buildings blocks in its project plan, standardizing the data flow over the 9 hospitals. The dataflow typically starts uploading raw data from source databases via a data gateway into a storage component that does pseudonimization steps-if relevant- and stores data in parquet file format. The raw data in parquet format is being ingested into a raw database via the ETL tool, then processed in a work database where it is cleaned and transformed according to the mapping files. Finally, the data is standardized into the OMOP format in an OMOP database ready for analysis and benchmarking.

Each hospital will set up a production proof infrastructure, either cloud or on premise components are chosen.
A data gateway will enable hospitals to schedule and execute tasks that either can upload data from source databases towards the cloud or an on-premise storage component, upload vocabularies and mappings, do pseudonimization jobs, cast raw data in parquet files ready for ETL query jobs.
RiAB® ETL tool is able to transform electronic medical records into the OMOP common data model. It automatically extract data from source databases, transforms data into OMOP concepts and populates OMOP tables via simple CLI commands following the OMOP CDM 5.4 ETL conventions.
In data preparation, all partners employ a mapping tool named KEUN to adhere to established mapping guidelines. This process involves creating files that link source values to standard concepts, utilizing standardized vocabularies. These mappings facilitate the integration of disparate coding systems into a unified data model.
After in-house OMOP databases are populated, a data exchange technology can share insights between hospitals. A federated technology involves a system where multiple separate OMOP databases can be queried, analyzed, or utilized in a coordinated manner while the data itself remains in its original hospital database location. A FHIR proof-of-concept is set up illustrating a use case sharing data in a consistent format between hospitals.
Data quality & demo (analytics) 2 AI use cases on lung and prostate cancer are chosen as demonstrators besides a benchmarking dashboard.

Impact and future directions

Ready to tack any research and data-driven question on a project basis and truely collaborate between hospitals and operationalise these solutions in practice.

General info and contact

Keywords (#): OMOP, harmonizing, federated learning, AI

Website: www.fhin.be

RADar project research lead: Ir. Kim Denturck

RADar project researchers: Ir. Kim Denturck, B.Sc. Pieter-Jan Lammertyn, dr. Siel Depestele, M.Sc. Louise Berteloot,
Dr. Nathalie Mertens, Prof. Dr. Ir. Peter De Jaeger

Principal investigator: To be determined

Timeline: July 2023 – December 2025

Status: Ongoing

FHIN® Hospital Partners:
AZ Delta, AZ Klina, AZ Sint-Jan, CHU Liège, Imelda Ziekenhuis, UZ Antwerpen, UZ Brussel, UZ Gent,
Ziekenhuis Oost-Limburg

Funding: via FOD data capabilities call

24 October 2024